a  di\ @s:ddlmZmZmZmZddlZddlmZmZddlm Z ddl m Z m Z m Z mZmZmZmZmZmZmZmZmZddlmZmZmZddlmZmZddlmZdd l m!Z!dd l"m#Z#d2eeeefeeeefee d eeed ddZ$d3ee%eee&ee%efee&ddddZ'd4eeeeeffeeee%e&edddZ(d5eeee%eee&ee%efee&e)edddZ*d6e&ee d eee&ee%efee&ddddZ+d7eeeeeffe&ee d eeeddd Z,d8eee&ee d eee&ee%efee&e)ed!d"d#Z-d9e&ee d$eee&ee%efee&dd%d&d'Z.d:eeeeeffe&ee d$eeee&ed(d)d*Z/d;eee&ee d$eee&ee%efee&e)ed+d,d-Z0d7z!_reduce_auroc..NrzUAverage precision score for one or more classes was `nan`. Ignoring these classes in z-averagerrzBReceived an incompatible combinations of inputs to make reduction.) isinstancerrtorchstackzipisnananyr UserWarningmeanrsum ValueError)r r!r"r#residxr*r*r+ _reduce_auroc-s"   r:)max_fpr thresholds ignore_indexr$cCsFt|||durBt|tsBd|kr0dkrBnntd|dS)Nrr&z@Arguments `max_fpr` should be a float in range (0, 1], but got: )r r.floatr7)r;r<r=r*r*r+_binary_auroc_arg_validationHs *r?r&)stater<r; pos_labelr$cCs$t|||\}}}|dus:|dks:|dks:|dkrFt||dSt|trV|jn|dj}t||d}tj||ddd} ||| d|| || d} t || d|| | } t |d| | dg}t |d| | dg}t||d} d|d} dd| | || S) Nr&rr%)deviceT)Z out_int32rightg?) rr6rr.rrBrr/Z bucketizeZlerpcatview)r@r<r;rAr r!_Z_deviceZmax_areastopZweightZ interp_tprZ partial_aucZmin_arear*r*r+_binary_auroc_computeRs(  $  rIT)predstargetr;r<r= validate_argsr$cCsH|rt|||t|||t||||\}}}t|||}t|||S)a Compute Area Under the Receiver Operating Characteristic Curve (`ROC AUC`_) for binary tasks. The AUROC score summarizes the ROC curve into an single number that describes the performance of a model for multiple thresholds at the same time. Notably, an AUROC score of 1 is a perfect score and an AUROC score of 0.5 corresponds to random guessing. Accepts the following input tensors: - ``preds`` (float tensor): ``(N, ...)``. Preds should be a tensor containing probabilities or logits for each observation. If preds has values outside [0,1] range we consider the input to be logits and will auto apply sigmoid per element. - ``target`` (int tensor): ``(N, ...)``. Target should be a tensor containing ground truth labels, and therefore only contain {0,1} values (except if `ignore_index` is specified). The value 1 always encodes the positive class. Additional dimension ``...`` will be flattened into the batch dimension. The implementation both supports calculating the metric in a non-binned but accurate version and a binned version that is less accurate but more memory efficient. Setting the `thresholds` argument to `None` will activate the non-binned version that uses memory of size :math:`\mathcal{O}(n_{samples})` whereas setting the `thresholds` argument to either an integer, list or a 1d tensor will use a binned version that uses memory of size :math:`\mathcal{O}(n_{thresholds})` (constant memory). Args: preds: Tensor with predictions target: Tensor with true labels max_fpr: If not ``None``, calculates standardized partial AUC over the range ``[0, max_fpr]``. thresholds: Can be one of: - If set to `None`, will use a non-binned approach where thresholds are dynamically calculated from all the data. Most accurate but also most memory consuming approach. - If set to an `int` (larger than 1), will use that number of thresholds linearly spaced from 0 to 1 as bins for the calculation. - If set to an `list` of floats, will use the indicated thresholds in the list as bins for the calculation - If set to an 1d `tensor` of floats, will use the indicated thresholds in the tensor as bins for the calculation. ignore_index: Specifies a target value that is ignored and does not contribute to the metric calculation validate_args: bool indicating if input arguments and tensors should be validated for correctness. Set to ``False`` for faster computations. Returns: A single scalar with the auroc score Example: >>> from torchmetrics.functional.classification import binary_auroc >>> preds = torch.tensor([0, 0.5, 0.7, 0.8]) >>> target = torch.tensor([0, 1, 1, 0]) >>> binary_auroc(preds, target, thresholds=None) tensor(0.5000) >>> binary_auroc(preds, target, thresholds=5) tensor(0.5000) )r?r r r rI)rJrKr;r<r=rLr@r*r*r+ binary_aurocms ?   rM) num_classesr"r<r=r$cCs0t|||d}||vr,td|d|dS)N)rrrN)Expected argument `average` to be one of but got )r r7)rNr"r<r=allowed_averager*r*r+ _multiclass_auroc_arg_validations rR)r@rNr"r<r$c CsZt|||\}}}t||||dur6t|d|dn|ddddddfddS)Nr&)Z minlengthrr#)rr:rr>r6)r@rNr"r<r r!rGr*r*r+_multiclass_auroc_computes:rU)rJrKrNr"r<r=rLr$cCsR|r t||||t||||t|||||\}}}t||||}t||||S)aCompute Area Under the Receiver Operating Characteristic Curve (`ROC AUC`_) for multiclass tasks. The AUROC score summarizes the ROC curve into an single number that describes the performance of a model for multiple thresholds at the same time. Notably, an AUROC score of 1 is a perfect score and an AUROC score of 0.5 corresponds to random guessing. Accepts the following input tensors: - ``preds`` (float tensor): ``(N, C, ...)``. Preds should be a tensor containing probabilities or logits for each observation. If preds has values outside [0,1] range we consider the input to be logits and will auto apply softmax per sample. - ``target`` (int tensor): ``(N, ...)``. Target should be a tensor containing ground truth labels, and therefore only contain values in the [0, n_classes-1] range (except if `ignore_index` is specified). Additional dimension ``...`` will be flattened into the batch dimension. The implementation both supports calculating the metric in a non-binned but accurate version and a binned version that is less accurate but more memory efficient. Setting the `thresholds` argument to `None` will activate the non-binned version that uses memory of size :math:`\mathcal{O}(n_{samples})` whereas setting the `thresholds` argument to either an integer, list or a 1d tensor will use a binned version that uses memory of size :math:`\mathcal{O}(n_{thresholds} \times n_{classes})` (constant memory). Args: preds: Tensor with predictions target: Tensor with true labels num_classes: Integer specifing the number of classes average: Defines the reduction that is applied over classes. Should be one of the following: - ``macro``: Calculate score for each class and average them - ``weighted``: calculates score for each class and computes weighted average using their support - ``"none"`` or ``None``: calculates score for each class and applies no reduction thresholds: Can be one of: - If set to `None`, will use a non-binned approach where thresholds are dynamically calculated from all the data. Most accurate but also most memory consuming approach. - If set to an `int` (larger than 1), will use that number of thresholds linearly spaced from 0 to 1 as bins for the calculation. - If set to an `list` of floats, will use the indicated thresholds in the list as bins for the calculation - If set to an 1d `tensor` of floats, will use the indicated thresholds in the tensor as bins for the calculation. ignore_index: Specifies a target value that is ignored and does not contribute to the metric calculation validate_args: bool indicating if input arguments and tensors should be validated for correctness. Set to ``False`` for faster computations. Returns: If `average=None|"none"` then a 1d tensor of shape (n_classes, ) will be returned with auroc score per class. If `average="macro"|"weighted"` then a single scalar is returned. Example: >>> from torchmetrics.functional.classification import multiclass_auroc >>> preds = torch.tensor([[0.75, 0.05, 0.05, 0.05, 0.05], ... [0.05, 0.75, 0.05, 0.05, 0.05], ... [0.05, 0.05, 0.75, 0.05, 0.05], ... [0.05, 0.05, 0.05, 0.75, 0.05]]) >>> target = torch.tensor([0, 1, 3, 2]) >>> multiclass_auroc(preds, target, num_classes=5, average="macro", thresholds=None) tensor(0.5333) >>> multiclass_auroc(preds, target, num_classes=5, average=None, thresholds=None) tensor([1.0000, 1.0000, 0.3333, 0.3333, 0.0000]) >>> multiclass_auroc(preds, target, num_classes=5, average="macro", thresholds=5) tensor(0.5333) >>> multiclass_auroc(preds, target, num_classes=5, average=None, thresholds=5) tensor([1.0000, 1.0000, 0.3333, 0.3333, 0.0000]) )rRrrrrU)rJrKrNr"r<r=rLr@r*r*r+multiclass_aurocsN rV)microrrr) num_labelsr"r<r=r$cCs0t|||d}||vr,td|d|dS)N)rWrrrNrOrP)rr7)rXr"r<r=rQr*r*r+ _multilabel_auroc_arg_validation's rY)r@rXr"r<r=r$c Cs|dkr|t|tr.|dur.t|d|ddS|d}|d}|durj||k}||}||}t||f|ddSt||||\}} } t|| ||dur|ddkjddn|ddddddfddS)NrWr&)r;r)ZdimrSrT)r.rrIr6flattenrr:r>) r@rXr"r<r=rJrKr9r r!rGr*r*r+_multilabel_auroc_compute3s"    >r[)rJrKrXr"r<r=rLr$cCsT|r t||||t||||t|||||\}}}t||||}t|||||S)akCompute Area Under the Receiver Operating Characteristic Curve (`ROC AUC`_) for multilabel tasks. The AUROC score summarizes the ROC curve into an single number that describes the performance of a model for multiple thresholds at the same time. Notably, an AUROC score of 1 is a perfect score and an AUROC score of 0.5 corresponds to random guessing. Accepts the following input tensors: - ``preds`` (float tensor): ``(N, C, ...)``. Preds should be a tensor containing probabilities or logits for each observation. If preds has values outside [0,1] range we consider the input to be logits and will auto apply sigmoid per element. - ``target`` (int tensor): ``(N, C, ...)``. Target should be a tensor containing ground truth labels, and therefore only contain {0,1} values (except if `ignore_index` is specified). Additional dimension ``...`` will be flattened into the batch dimension. The implementation both supports calculating the metric in a non-binned but accurate version and a binned version that is less accurate but more memory efficient. Setting the `thresholds` argument to `None` will activate the non-binned version that uses memory of size :math:`\mathcal{O}(n_{samples})` whereas setting the `thresholds` argument to either an integer, list or a 1d tensor will use a binned version that uses memory of size :math:`\mathcal{O}(n_{thresholds} \times n_{labels})` (constant memory). Args: preds: Tensor with predictions target: Tensor with true labels num_labels: Integer specifing the number of labels average: Defines the reduction that is applied over labels. Should be one of the following: - ``micro``: Sum score over all labels - ``macro``: Calculate score for each label and average them - ``weighted``: calculates score for each label and computes weighted average using their support - ``"none"`` or ``None``: calculates score for each label and applies no reduction thresholds: Can be one of: - If set to `None`, will use a non-binned approach where thresholds are dynamically calculated from all the data. Most accurate but also most memory consuming approach. - If set to an `int` (larger than 1), will use that number of thresholds linearly spaced from 0 to 1 as bins for the calculation. - If set to an `list` of floats, will use the indicated thresholds in the list as bins for the calculation - If set to an 1d `tensor` of floats, will use the indicated thresholds in the tensor as bins for the calculation. ignore_index: Specifies a target value that is ignored and does not contribute to the metric calculation validate_args: bool indicating if input arguments and tensors should be validated for correctness. Set to ``False`` for faster computations. Returns: If `average=None|"none"` then a 1d tensor of shape (n_classes, ) will be returned with auroc score per class. If `average="micro|macro"|"weighted"` then a single scalar is returned. Example: >>> from torchmetrics.functional.classification import multilabel_auroc >>> preds = torch.tensor([[0.75, 0.05, 0.35], ... [0.45, 0.75, 0.05], ... [0.05, 0.55, 0.75], ... [0.05, 0.65, 0.05]]) >>> target = torch.tensor([[1, 0, 1], ... [0, 0, 0], ... [0, 1, 1], ... [1, 1, 1]]) >>> multilabel_auroc(preds, target, num_labels=3, average="macro", thresholds=None) tensor(0.6528) >>> multilabel_auroc(preds, target, num_labels=3, average=None, thresholds=None) tensor([0.6250, 0.5000, 0.8333]) >>> multilabel_auroc(preds, target, num_labels=3, average="macro", thresholds=5) tensor(0.6528) >>> multilabel_auroc(preds, target, num_labels=3, average=None, thresholds=5) tensor([0.6250, 0.5000, 0.8333]) )rYrrrr[)rJrKrXr"r<r=rLr@r*r*r+multilabel_aurocOsR r\)binaryZ multiclassZ multilabel) rJrKtaskr<rNrXr"r;r=rLr$c Cst|}|tjkr&t|||||| S|tjkrbt|tsNtdt|dt ||||||| S|tj krt|tstdt|dt ||||||| SdS)aCompute Area Under the Receiver Operating Characteristic Curve (`ROC AUC`_). The AUROC score summarizes the ROC curve into an single number that describes the performance of a model for multiple thresholds at the same time. Notably, an AUROC score of 1 is a perfect score and an AUROC score of 0.5 corresponds to random guessing. This function is a simple wrapper to get the task specific versions of this metric, which is done by setting the ``task`` argument to either ``'binary'``, ``'multiclass'`` or ``multilabel``. See the documentation of :func:`~torchmetrics.functional.classification.binary_auroc`, :func:`~torchmetrics.functional.classification.multiclass_auroc` and :func:`~torchmetrics.functional.classification.multilabel_auroc` for the specific details of each argument influence and examples. Legacy Example: >>> preds = torch.tensor([0.13, 0.26, 0.08, 0.19, 0.34]) >>> target = torch.tensor([0, 0, 1, 1, 1]) >>> auroc(preds, target, task='binary') tensor(0.5000) >>> preds = torch.tensor([[0.90, 0.05, 0.05], ... [0.05, 0.90, 0.05], ... [0.05, 0.05, 0.90], ... [0.85, 0.05, 0.10], ... [0.10, 0.10, 0.80]]) >>> target = torch.tensor([0, 1, 1, 2, 2]) >>> auroc(preds, target, task='multiclass', num_classes=3) tensor(0.7778) z+`num_classes` is expected to be `int` but `z was passed.`z*`num_labels` is expected to be `int` but `N) rZfrom_strBINARYrMZ MULTICLASSr.intr7typerVZ MULTILABELr\) rJrKr^r<rNrXr"r;r=rLr*r*r+aurocs)      rb)rN)NNN)Nr&)NNNT)rNN)rN)rNNT)NN)N)rNNT)NNNrNNT)2typingrrrrr/rrZtyping_extensionsrZ=torchmetrics.functional.classification.precision_recall_curver r r r r rrrrrrrZ*torchmetrics.functional.classification.rocrrrZtorchmetrics.utilities.computerrZtorchmetrics.utilities.datarZtorchmetrics.utilities.enumsrZtorchmetrics.utilities.printsrr:r>r`r?rIboolrMrRrUrVrYr[r\rbr*r*r*r+s  8        I      [      `